Abstract

Administrative data are increasingly central to the production of official labour market and migration statistics and are often treated as a benchmark for data quality. This paper examines the quality of nationality information in German administrative labour market data by exploiting a rare linkage between survey and administrative sources. We use data from a the large-scale online migration survey “International Mobility Panel of Migrants in Germany” (IMPa), linked at the individual level to administrative labour market records from the “Integrated Employment Biographies” (IEB) that merges various administrative data sources to create detailed, day-by-day records of individuals' labor market histories in Germany. Assuming that survey-reported nationality is correct, we assess the extent, persistence, and determinants of discrepancies between nationality information recorded in the survey and in administrative sources.
We distinguish four types of classification outcomes: correctly classified individuals, Germans misclassified as foreigners, foreigners misclassified as Germans, and foreigners with an incorrect nationality recorded. These error types are analysed overall and separately by administrative source, distinguishing employer reports from records generated by job centres and employment agencies. In addition, we examine the contexts in which nationality errors are most likely to occur and how long they persist over time.
In a second step, we assess the quality of administrative metadata by analysing whether the timing of first occurrences in administrative records can be used as proxies for key migration-related events reported in the survey. Specifically, we study the time lag between survey-reported entry into Germany and the first administrative notification, as well as between the reported year of acquisition of German citizenship and the first occurrence of German nationality in administrative data. We document substantial heterogeneity across countries of origin and socio-demographic groups, highlighting conditions under which such metadata-based approximations are more or less reliable.
Finally, we propose a practically implementable imputation strategy that exploits differences in reporting quality across administrative sources. By prioritising higher-quality sources and correcting implausible isolated changes in nationality, the approach substantially reduces misclassification rates. Our findings demonstrate that administrative data quality varies systematically by source and institutional context and that targeted quality management strategies can significantly improve the reliability of key variables. The results have direct implications for the production, interpretation, and quality assurance of migration-related indicators based on administrative data in official statistics.